Dr Paul Yoo

Overview

Biography

Prior to his current post at Birkbeck (Univ. of London), Paul has held academic and research positions at esteemed institutions such as Cranfield (Defence Academy of the UK), Sydney (USyd) and South Korea (KAIST). He was trained originally as a data scientist with degrees from the University of Sydney, Australia, and has since published over one hundred papers in prestigious journals and conferences. He has also overseen over US$2.5 million in project funding as the PI and has received various national and international awards for his work in advanced data analytics, machine learning, and secure systems research. These include the IEEE Outstanding Leadership Award, Rozetta Award (formerly CMCRC), Emirates Foundation Research Award, and the ICT Fund Award. Most recently, he was awarded the Samsung award for his research on novel abstraction machine-learning approach [news], Research England’s Global Challenge Research Fund (GCRF), as well as the MPU-Macau Fund.

Paul currently serves as an Editor for ACM Computing Surveys (Q1). He has previously served as an Editor for IEEE Transactions on Sustainable Computing (Q1), IEEE COMML (Q1) and IEEE Access (Q1). He is also affiliated with the University of Sydney and Korea Advanced Institute of Science and Technology (KAIST) as a Visiting Professor. Paul is a Senior Member of the IEEE and a Fellow of HEA.

As the Founder and Chair of BIDA's Threat Intelligence Lab, Paul leads a team dedicated to merging advancements in machine learning with the evolution of cybersecurity. The lab is at the forefront of developing the next generation of intelligent cyber defence mechanisms. It fosters a cutting-edge security research environment that focuses on both reactive detection and proactive prediction of threats. His team has pioneered big data-based proactive forecasting methods for assisting strategic descision makers. The lab currently houses 11 PhD students, and recent photos of their lab meetings can be found here.

Paul’s research is centred around the theory and methodology of machine learning for large-scale real-world problems. He has successfully applied various machine learning and big data analytic approaches to diverse problem domains, such as psychology, security, biology, finance, and manufacturing.

Paul has also supervised a number of PhD/MSc by Research students to completion, and he welcomes inquiries about research degree (PhD/MPhil) and internship opportunities.

Qualifications

• PhD in Engineering, University of Sydney, Australia, 2009
• PgCert in Academic Practice, Cranfield University, 2019

Web profiles

• Personal Web Profile

Administrative responsibilities

• Dy Director (Knowledge Exchange), Birkbeck Institute for Data Analytics (BIDA)
• Head, Data Science and AI Group
• Chair, Threat Intelligence Lab
• Programme Director, Digital Tech Solutions (Software)

Professional memberships

• Senior Member, IEEE

• Fellow, HEA

Honours and awards

• Samsung Global Research Outreach Award, Samsung, January 2017
• IEEE Outstanding Leadership Award, IEEE, January 2013
• Rozetta Award, Rozetta Institute, Australia, July 2006

ORCID

0000-0001-7665-8616

Supervision and teaching

Supervision

Supervisions within Birkbeck:

12

Current doctoral researchers

• THIAGO SUZUKI

  Nature of supervision:

  Principal Supervisor, Oct 2022 onwards

• ALBERTO MATUOZZO

  Nature of supervision:

  Principal Supervisor, Oct 2020 onwards

  Title of thesis:

  Machine Learning techniques for equity markets: from custom loss functions in predictive modelling, to dynamic network ensemble learning for capital deployment.

  Description:

  Machine learning techniques are the underpinning of the transition to data driven finance: the aim is not to replace economic theory or corporate finance, but rather enhance them, deploying models that can learn from non linear, discontinuous or hierarchical data structures and that can foster the engineering of features that could not be observed directly.

  This research explores an area at the intersection of finance and computer science, investigating machine learning techniques for equity markets prediction, processing signals from different set of features (statistical, technical, fundamental and sentiment). The aim is to research the development and application of techniques that can have an impact in the financial industry.

  This research aims to propose solutions to tackle the problems of equity market forecasting given the fact that securities prices undergo different regimes, and price formation is affected by several factors of different nature (micro, macro, sentiment).

• DILARA YAZAN

  Nature of supervision:

  Principal Supervisor, Jan 2021 onwards

• GASSO MWALUSEKE

  Nature of supervision:

  Principal Supervisor, Oct 2020 onwards

• GURMENDER SINGH ATWAL

  Nature of supervision:

  Principal Supervisor, Jan 2020 onwards

• LAWRENCE OLUSANYA

  Nature of supervision:

  Principal Supervisor, Oct 2020 onwards

• MARIO RODRIGUEZ URETA

  Nature of supervision:

  Principal Supervisor, Oct 2023 onwards

  Title of thesis:

  Digital Twins of Macroeconomic Models

  Description:

  The objective of this research is to introduce an innovative macroeconomic digital twin framework intended for policymakers. By constructing a twin digital  of real-world economic phenomena, we can  analyse the potential impact of macroeconomic policies within a digital environment before their actual implementation in reality. Our contingency  is based on the recognition that forecasting in social sciences, particularly in economics, often needs a reasonable degree of accuracy in predicting events. The lack of accuracy in forecasting in economics is primarily due to the underutilization of appropriate statistical and computational techniques. Instead, policymakers frequently rely on methods that often yield impractical outcomes. Leveraging twin technologies can be a valuable tool, empowering policymakers to explore and assess various scenarios effectively before implementation.  

  The aim of the research is to:  

  “Create a twin digital of an economic system (society, country, or industry) that allow us to simulate and reproduce different economic policies implementation and outcomes”. 

  The objectives of the research are:  

  • To review twin technologies methodologies, policy making theories, forecasting methods in economics, and behavioural economics.  
  • To create a twin model of an economic system.
  • To create an agent based model in Python  that will be the foundation  of the twin model. 
  • To validate our model using historical data and traditional statistical tools such r square, standard error and mean average performance error (MAPE).   

  • Agent Based Model (ABM) will be the technology chosen to create the digital twin simulation. ABMs are able to create intricate social systems, incorporating different agents and allowing these agents to dynamically change. This in turn, allow us to simulate different social contexts

  After reviewing existing literature, we define the components of our ABM,  

  • The agent/s. This can be one ore more agents. For instance, if the aim is to replicate a entire society, the agents could  be the public sector, the private sector, households, and international relations.   
  • The agent/s rules. These rules must be based on behavioural economics and general economic principles. 
  • The environment.  The environment will be the economic ecosystem. For instance, if the model aims to replicate a specific country then variables like  the base level of pricing, debt, interest rates and other macroeconomic variables must be set prior to run the simulation.  
  • The environment rules. Regulatory policies, regulations and economic principles that dictate the how these variables act.  Decision rules. Once the agent and the environment is defined, then the decision rules that trigger behaviour and the magnitude of the effect must be defined. Here we also define loops and reactions.  

  Once the ABM is created and in working conditions, we validate our model by using historical data. For instance, we can validate our simulation by forecasting  GDP growth and running the model under different scenarios (actual vs forecasted) and compare the outcome and the performance against historical models.  

• NEIL MACKINNON

  Nature of supervision:

  Principal Supervisor, Oct 2020 onwards

• OMAR ALHAWI

  Nature of supervision:

  Principal Supervisor, Oct 2023 onwards

  Title of thesis:

  Securing mobile Devices through Advanced Anomaly Detection: A Study on Android and iOS

Doctoral alumni since 2013-14

• ZAID ALMAHMOUD

  Nature of supervision:

  Principal Supervisor, Oct 2021

  Title of thesis:

  Forecasting Cyber Threats & Pertinent Alleviation Technologies

  Awarded:

  2024

• SEONGIL HAN

  Nature of supervision:

  Joint Principal Supervisor, Jun 2020

  Title of thesis:

  Explainable Credit Scoring through Generative Adversarial Networks

  Awarded:

  2021

Teaching

Teaching modules

• Foundations of Data Science II (BUCI070H5)
• Applied Machine Learning (BUCI077H7)

Publications